question the mere face validity of this instrument. Bambini and associates (2009) and Goldenberg and associates (2005) also used new instruments but psychometric properties were determined and found to be acceptable, with Cronbach’s α = 0.817 (pre-test) and 0.858 (post-test) (Bambini et al., 2009) and 0.97 (Goldenberg et al., 2005). Both of the latter studies employed SCT as a guiding theoretical framework and used pre-test/post-test designs,
supporting assertions that the intervention (simulation) affected the outcome variable, self- efficacy. The principal limitation of each of these studies was the lack of use of a control group. Without a control group, improvement in self-efficacy could have occurred for various reasons, including subject maturation. It is also possible that participants would have reported increased self-efficacy with the standard education, rather than the simulation intervention.
Innovations of this Study
There is a paucity of interventional research focused on LV and no published studies preparing nursing students for the LV they are likely to encounter in the hospital work setting both as students and newly licensed nurses. In addition, though self-efficacy is mentioned in nursing literature with regard to LV, no studies have measured self-efficacy as an outcome variable using a validated and reliable instrument. Despite the limitations of previous studies, as discussed in both Chapters 2 and 3, cognitive-behavioral rehearsal implemented as a simulation holds potential for the purpose of this study. This study adds to the existing body of knowledge by: (1) it measures self-efficacy in relation to LV response, a construct which has not been quantitatively measured in previous studies, (2) it was guided closely by a theoretical framework to ensure accuracy in variables tested, and (3) it used a more rigorous research design than have been implemented previously, yielding more reliable results,
20
reducing risk of both Type I and Type II errors, and providing sound basis for future educational interventions.
21 CHAPTER 3
METHODS Social Skills Training
Despite design flaws in previous studies, response rehearsal, a form of Cognitive Behavioral Therapy (CBT) holds potential as an effective intervention. CBTs are predicated upon Social Cognitive Theory, of which self-efficacy is an essential construct (Bandura, 1997). Social Skills Training (SST), a specialized form of CBT, has been used successfully with individuals who have not developed communication responses or had practice with more complex social interactions to develop appropriate communication patterns (Strong Kinneman & Bellack, 2012; Twohig & Dehlin, 2012). Nursing students who have not yet been exposed to lateral violence (LV) have a deficit in exposure and practice in developing responses to lateral violence which can require more sophisticated communication patterns. This study adds to the existing body of knowledge by: (1) measuring self-efficacy in relation to lateral violence response, a construct which has not been quantitatively measured in previous studies, (2) it was guided closely by a theoretical framework to ensure accuracy in variables tested, and (3) it utilized a more rigorous research design than have been
implemented previously, yielding more reliable results, reducing risk of both Type I and Type II errors, and providing sound basis for future interventions.
Research Design
This research utilized a longitudinal, experimental, randomized cluster design.
Participants from two baccalaureate nursing programs were randomly assigned to clusters by school affiliation. One cluster received the intervention (intervention group); the other served as the control group (attention-control group). Clustering participants by school affiliation
22
reduced the risk of contamination between groups by ensuring that the intervention and control groups were as mutually exclusive as possible. Randomization of participants enhanced the rigor of this study by eliminating bias in group assignment based on attributes of groups or individuals within groups which could impact outcome variables (Polit & Beck, 2012, p. 206). Randomization in this study was determined by a coin flip which is
appropriate for two-group randomization. Outcome variables can also be influenced by performance bias, participants’ inherent desire to perform well (Polit & Beck, 2012, p. 210). This research also implemented a single-blind procedure, wherein the participants were unaware of whether they are in the intervention or control group thus reducing performance bias.
Participant Selection
Participants recruited from two faith-based, baccalaureate, pre-licensure nursing programs within the same urban setting. These study sites were selected based on program attributes and availability, in order to maximize homogeneity among participants.
Participants were recruited by a member of the research team who was not responsible for course content, assigning grades, or present during the intervention, during their regularly scheduled class time. Inclusion criteria for participation included membership in the senior classes of two investigator-selected nursing programs, ability to read and write in English, and attendance in class on recruitment days. Students enrolled in an Advanced-Track (AT) program were excluded from recruitment. AT programs allow students holding a previously earned baccalaureate degree to complete the nursing program in a condensed amount of time. Due to this difference in educational background, AT students may have differing
23
characteristics from students enrolled in traditional programs which, in turn, may have influenced all variables being measured in this study.
Sample and Recruitment
Convenience sampling was utilized to recruit participants for this study. Students attending class on recruitment days had an equal opportunity to participate. Convenience sampling can introduce bias into studies, since those who choose to participate may do so based on a particular set of personal attributes (Polit & Beck, 2012, p. 276). However, convenience sampling is economical and an effective method of maximizing participation. Thus, it was considered to be appropriate for this research.
The recruiter spent 15 minutes at the beginning of class time to discuss the purpose of the study, the participant role, and to review the consent form (Appendix B; Appendix C). Consent forms were provided to all students; study instructions guided students who wished to participate to complete the consent form. Participants had an opportunity to ask questions both at this time and at time of the intervention. Consent forms were signed and collected on the day of distribution; however, participants were given the option to review the consent form and submit it two weeks later at the time of the intervention. Each participant also developed a unique study-specific password. This password allowed pre and post-test data to be correlated, eliminating collection of any identifying data and maintaining confidentiality to the responses. A copy of the consent form was provided for their personal files. An electronic study file was developed by the Principal Investigator (PI) for the purposes of linking the participant to their study number. This file was be maintained on the PI’s password-protected personal computer.
24 Sample size.
Apriori power analysis indicated that 32 participants in each cluster were sufficient to achieve a power of 0.80 with a moderate effect size of 0.35 (Cohen, 1988, p. 311). A total of 41 participants were recruited from the intervention site and 47 participants from the
attention-control site, for a total N = 88. This participation was sufficient to meet the requirements of power and effect size. The instrument used in this study contains 10 items and was used in both pre and post-testing, necessitating only 25 participants per cluster in order to validate results.
Human Subjects Considerations
This research underwent a full review and approval process by the Institutional Review Board (IRB) at the PI’s University. Following this approval, IRB approval at each of the two study sites was secured. These IRBs routinely oversee nursing research to ensure the ethical treatment of human subjects. The IRB at the PI’s University also requires all
investigators to be certified in CITI© (CITI Program, 2012) training. The CITI program is an agency which provides online training to investigators in biomedical and social sciences research.
Benefits to participants included: increased knowledge, ability, and self-efficacy in responding to LV. Indirect benefits, or benefits to society, included generating new scientific knowledge to help future nursing students and the nursing profession. The only foreseeable risk to participants was possible psychological distress related to distressing event recall, incurred by participation in emotionally-taxing role play scenarios. Participants were
instructed to report any distress during the intervention, whereupon the PI was to discontinue their participation and refer them to appropriate resources immediately. Counseling resources
25
are available at both study sites, at free or reduced rates for students, if participants had experienced psychological distress. During the intervention, no participants either reported or exhibited signs of distress.
Materials
Participants in the intervention group received printed materials, containing the dialogue to be practiced during the intervention (Appendix D). Participants in the attention- control group received printed materials, containing a weekly schedule, list of weekly activities, and instructions for completing the schedule (Appendix E)
Instrument.
The dependent variable, self-efficacy, was measured using an adaptation of the Scale to Address Disruptive Physician Behavior© (SADBS; Saxton, 2010). This scale was
previously used to measure peri-operative nurses’ self-efficacy in addressing disruptive physician behavior. Factor analysis was performed on this scale to establish its psychometric properties. Content validity was confirmed and a Cronbach’s α = 0.904 indicated excellent reliability (Saxton, 2010, p. 48). The SADBS© scale includes 10 items, measured on a 10- point Likert-type scale, asking participants to rate their perceived self-efficacy in responding to specific disruptive physician behaviors. Scale steps are arranged in increasing order such that 0 = not confident to 10 = highly confident. Summed scores using this instrument range from 0 - 100. For this research, the SADBS© was adapted by replacing the item stems so that participants were asked how confident they felt in responding to the 10 most common LV behaviors (SADBS-R; Appendix B). Permission to adapt and use the SADBS was obtained (Appendix A).
26
The fourth assumption of this study was that participants would respond to the instrument items honestly. Violation of this assumption would introduce internal bias into the study; thus it was important to include a social desirability item. This item was selected from a list of such items by Crowne and Marlow (1960) and was considered particularly
appropriate for its content. The social desirability item read “I have never deliberately done or said something to hurt someone’s feelings” and participants were asked to rate their confidence in this statement on the same 0 – 10 Likert-type scale.
Demographic data collected including age, gender (M/F), previous experience with/exposure to LV (Y/N), and previous education on bullying (Y/N) was reported in aggregate form to describe the population. Previous exposure to/experience with LV and previous training regarding workplace bullying were accounted for as possible covariates during data analysis.
Procedures
One cluster received the intervention; the other served as the control group, thereby enhancing the rigor of the study. Clustering also reduced the risk of participant contamination by ensuring that the intervention and control groups are as mutually exclusive as possible. Clusters were randomly assigned to either treatment or attention control by coin flip.
Intervention Group: At the beginning of the intervention for the intervention group the recruiter explained the study, including the purpose, time requirement, and data to be collected. Each participant then completed the SADBS-R pre-test and provided demographic information. Once all data were collected, the intervention was provided. At the conclusion of the intervention, each participant completed the SADBS-R, providing the first set of post- test data. Participants also completed the SADBS-R three months after the intervention to
27
assess for longitudinal effects of the intervention on self-efficacy to respond to LV. Participants in the intervention group received a one-hour SST intervention, aimed at developing appropriate and effective responses to LV (Appendix E). This aim was achieved by (1) modeling, (2) role play, and (3), feedback, which are the essential steps of SST. A guided discussion followed the intervention, allowing these participants to describe this experience.
Attention Control Group: At the beginning of the intervention for the control group, the PI explained the study, including the purpose, time requirement, and data to be collected. Each participant then completed the SADBS-R pre-test and provided demographic
information. Once all data were collected, the intervention was provided. At the conclusion of the intervention, participants completed the SADBS-R, providing the first set post-test data for this group. Participants in this group also completed the SADBS-R three months after the intervention to assess for longitudinal effects of the intervention on self-efficacy to respond to LV. Participants in this cluster received a one-hour intervention focused on time management as a stress-reduction technique (Appendix F).
Data Collection
Study data included pre and post-test questionnaires, with each study cluster
separately maintained. Pre-tests were completed by participants and collected by the PI prior to each intervention. Post-tests were completed by participants and collected by the PI following the conclusion of the intervention. Participants from both clusters also completed the SADBS-R three months after the intervention to assess for longitudinal effects of the intervention on self-efficacy to respond to LV. Scores from the SADBS-R were entered into a cluster-specific database on the PI’s personal password-protected computer. Pre and post-
28
test data were linked, using participants’ self-created identifiers. These identifiers were not linked to individual participants.
29 CHAPTER 4 DATA ANALYSIS
There is a paucity of research testing interventions aimed at reducing lateral violence (LV) and no published studies which focus on educating nursing students about appropriate responses. This scenario results in disadvantages and an inability to respond to the workplace violence that will be encountered as a newly graduated Registered Nurse (RN).
A total of 88 participants completed all study activities; 41 in the intervention group and 47 in the control group. The responses from these participants were hand-entered into study-specific SPSS files and triple-checked for accuracy. All statistical analyses were performed using SPSS 20.0. All demographic items were completed. No study survey item displayed missing data. There were four instances of failure to reply to the social desirability item. The social desirability item was added to the instrument to identify participant bias but was not intended as part of the study data set. Thus, substitute calculation for this data was not performed. Summed scored for individual responses on the SADBS-R were calculated and added to each data set. Higher summed scores reflect greater self-efficacy in responding to LV.
Descriptive Statistics Intervention Group
Descriptive statistical techniques were used to describe each study population. There were 41 participants in the intervention group. All participants were female; 80.5% (n =33) were between the ages of 20-25 years; 12.2% (n =5) were between the ages of 26-30 years, 4.9% (n =2) were between the ages of 31-35 years; none were between the ages of 36-40 years; and 2.4% (n = 1) were 40 years or older. Previous experience with workplace bullying
30
was reported by 80.5% (n = 33) of participants yet only 1.5% (n = 7) reported having received training on workplace bullying.
Attention-Control Group
There were 47 participants in the attention-control group. Females comprised 91.5% (n = 43) of this group and males 8.5% (n = 4). Age distribution in this group was similar to the intervention group with the majority (78.7%) of participants between the ages of 20-25 years
(n = 37); 8.5% between the ages of 26-30 years (n = 4); 4.3% between the ages of 31-35 years
(n = 2); none between the ages of 36-40 years; and 8.5% of 41 years or older (n = 4). Among this group, 40.4% (n = 19) reported previous exposure to workplace bullying, while the remaining 59.6% (n = 28) had not, and 61.7% (n = 29) reported having received previous training about workplace bullying, while the remaining 38.3% (n = 18) had not.
Comparison
Age distribution was fairly homogenous between the intervention and attention- control group, with the majority of participants between the ages of 20-25 years. While the intervention group was all female, 8.5% (n = 4) of the attention-control group was male. The most striking difference between the two groups was with regard to previous exposure to and training about workplace bullying. A far smaller percentage of the attention-control group reported previous exposure to workplace bullying (40.4%; n = 19) as compared to the intervention group (80.5%; n = 33) yet a larger percentage reported having received prior training with regard to workplace bullying (61.7%; n = 29) as compared to the intervention group (17.5%; n = 7).
31
Measures of Central Tendency Intervention Group
Mean responses on the pre-test among the intervention group, for the items scaled between 0-10 were between 4.09 and 5.17. However, a wide variation in responses contributed to means, with three items ranging nine points and the remaining seven items ranging 10 points on the 0-10 point instrument scale. Standard deviations ranged from 2.35- 3.51 points. Post-test response means were higher than pre-test means, ranging between 6.70 and 7.69 on the 0-10 instrument scale with smaller standard deviations between 1.80 and 2.23 points. This overall increase in scores was reflected in smaller ranges of responses, with only three items receiving a full 10 point range on the post-test. All measures of central tendency for the pre-test and post-test are reported in Tables 1 and 2 below.
Table 1
32
Table 2
Intervention Group Post-test Measures of Central Tendency
Attention-Control Group
Mean responses on the pre-test among the attention-control group were between 6.10 and 8.10 on the 0 - 10 instrument scale. Variation in responses included three items ranging eight points and seven items ranging 10 points on the 0 - 10 point instrument scale. Standard deviations ranged from 2.39 and 3.32 points. Post-test response means were only slightly higher than pre-test means, ranging between 6.44 and 9.12 on the 0 - 10 point instrument scale with standard deviations between 2.04 and 2.67 points. Ranges of responses on the post-test were similar to those of the pre-test with two items ranging seven points, two items ranging eight points, three items ranging nine points, and three items ranging 10 points. All measures of central tendency for the pre-test and post-test are reported in Tables 3 and 4 below.
33
Table 3
Attention-Control Group Pre-Test Measures of Central Tendency
Table 4
Attention-Control Group Post-Test Measures of Central Tendency
Comparison
The attention-control group scored higher overall on both the pre and post-tests, yet the intervention group showed more significant increases in all measures of central tendency. The intervention group’s responses were also more normally distributed on both the pre and post-tests than those of the attention-control group. This difference in distribution may be partially accounted for by the relatively higher reported incidence of receiving prior training about workplace bullying by the attention-control group.
34
Instrument Reliability
The instrument used to measure participant responses in this study, the SADBS-R, is an adaptation of a previously validated SADBS ©. Previous factor analysis on the SABDS © indicated excellent reliability with a Cronbach’s α = 0.904 (Saxton, 2010).
Intervention group
Reliability on the SADBS-R was first examined by determining the effect of social desirability item on the instrument on overall reliability, using both pre-test and post-test responses. Including the social desirability item, the Cronbach’s α = 0.927. Without the social desirability item, the Cronbach’s α = 0.947 (Table 5). Thus, it was concluded that the participants had not responded in a socially desirable manner, eliminating concern of this bias. Next, reliability of the pre-test and post-test were examined separately. Pre-test
reliability was determined at a Cronbach’s α = 0.925 and post-test reliability was determined at Cronbach’s α = 0.937 on the SADBS-R, excluding social desirability items.
35
Table 6
Intervention Group Item Total Statistics
Table 7
36
Table 8
Intervention Group Post-Test Item Total Statistics
Attention-Control Group
Including the social desirability item, the Cronbach’s α = 9.50. Excluding the social desirability item, the Cronbach’s α = 0.963. It was determined that the participants in this group had also not responded in a socially desirable manner, eliminating concern of this bias. Pre-test reliability was determined at a Cronbach’s α = 9.22 and post-test reliability was